采用Fisher线性判别分析进行MEG信号的分类

脑磁图(MEG)具有比脑电(EEG)信号更高的时空分辨率，可以作为输入信号建立脑-机接口系统.提出一种脑磁图的特征提取和分类方法，首先对MEG信号进行预处理，然后提取时域特征，最后采用Fisher线性判别分析进行分类.将该算法用于2008年脑-机接口数据竞赛的数据集Ⅲ，该数据集为一个典型的采用MEG信号的脑-机接口系统.离线分析结果表明，该算法取得了很好的分类准确率，对两个测试者(S1和S2)的分类正确率分别为5946%和4324%.与其他方法相比，该方法简单有效，运算速度快，具有较高的参考价值.

Classification of MEG Signals Using Fisher Linear Discriminant Analysis

The magnetoenephalography(MEG)signals have higher spatiotemporal resolution than EEG signals， which can be used as input signals to build brain computer interface(BCI)system. Feature extraction and classification methods of the MEG signals were introduced. Firstly， the MEG signals were preprocessed， and then time domain features were extracted. Finally， Fisher linear discriminant analysis(LDA)was used to classify the MEG signals. This algorithm was used to the data set Ⅲ of 2008 BCI competition which was a typical MEG based BCI system. The off line analysis results showed that high classification accuracy of 5946% and 4324% for two subjects(subject S1 and subject S2)could be obtained using this proposed algorithm. This algorithm is more efficient and simpler than others， which can be regarded as a good reference.